Integrating Edge Computing and Machine Learning for Thermal Anomaly Detection: A Space System Architecture

Traditionally, raw data is transmitted back to Earth using a bent-pipe architecture in response to ground station requests. However, this architecture does not scale well as the volume of data grows. This paper proposes an edge computing system architecture and a framework for thermal anomaly detection for space systems to address the drawbacks of the bent-pipe architecture. In such a manner, a standardized platform is presented, and the hardware and software architecture are broken down into layers and functionalities so that it is compatible and expandable with currently existing platforms. This approach entails the utilization of an infrared camera as an edge-sensing component to generate thermal profiles of electronic circuits, coupled with a field-programmable gate array (FPGA) serving as an edge computing system for onboard data processing to address thermal anomaly detection.